基于移去-恢复法的局部似大地水准面精化模型对比研究

针对似大地水准面在精化过程中不同移去-恢复高程转换模型的适用性和选取问题，基于二次曲面和EGM2008重力场模型，分别构建了RBF神经网络、多面函数和Shepard等3种类型的移去-恢复模型，结合平原和高原山区两个工程实例，通过调整高程拟合点数目进行似大地水准面拟合与精度对比。结果表明，在平原地区，EGM2008-多面函数高程模型精度略优于其他模型；在高原山区，当拟合点数较少时，基于EGM2008的移去-恢复模型精度高于基于二次曲面的移去-恢复模型，其中，EGM2008-多面函数和EGM2008-Shepard较优，而随着拟合点数目的增加，二次曲面-Shepard高程转换模型的精度优于其他模型。

Analysis and Comparison of Local Quasi Geoid Refinement Model Based on Remove-Restore Method

We are concerned with the applicability and selection of different remove-restore elevation conversion models in the process of quasi geoid refinement. We construct RBF neural networks, polyhedral function and Shepard function based on quadratic surfaces, and an EGM2008 gravity field model. We apply these tools to two engineering examples in the plain and plateau mountains, in order to fit normal height and compare the accuracy of each model by adjusting the number of fitting points. Experiments show that in the plain area, the accuracy of the EGM2008-polyhedral functions elevation model is slightly better than the other five models; in the plateau mountainous area, when the number of fitting points is fewer, the accuracy of the remove-restore models based on EGM2008 is higher than the corresponding models based on quadratic surfaces, and the EGM2008- polyhedral function model and the EGM2008-Shepard model are superior. As the number of fitting points increases, the accuracy of the quadratic surfaces-Shepard transformation model is better than the other five models.